How Minuteory Transformed Meeting Accountability and Customer Focus

I spent more than two decades inside meetings, most of them from the management side of the table. Technical product manager at Samsung for six years, then Endava, then a management position growing HYCU's Belgrade delivery center from 18 to 65 people, and finally C level at LoginEKO, where I restructured a 160-person organization twice; the same pattern repeated. Smart people spend an hour in a room, real decisions get made, and a week later nobody can tell you who owned what. The decision evaporated the moment people stood up. Minutes, when they existed at all, were written by the most junior person present, days late, and read by no one. I realized the problem was never note-taking. It was accountability. If a meeting produces a decision without a named owner and a deadline, it produces nothing. I built Minuteory to make that impossible to hide. Every meeting comes out with decisions, open questions, and tasks assigned to actual names.

What worked, LinkedIn from my personal profile, in a first-person founder voice. Posts that open with a scene or an uncomfortable truth instead of a product pitch, no corporate polish, link only in the first comment. It consistently outperforms everything we publish from the company page. People buy from a founder who sounds like a person, not from a logo. What flopped, automated competitor comparison content. Our internal analytics flagged comparison posts as our highest performing format, so the content pipeline kept replicating it. Then it published claims about a competitor's language support that were outdated because the vendor had shipped updates and our pipeline had no verification layer. The highest performing format turned out to be the highest risk format, and automation scaled the error faster than it scaled the reach. Now no competitor claim goes out without live verification against the vendor's own primary sources, with a 30-day staleness window. Painful lesson, cheap compared to what it could have cost.

Transcription quality in Serbian and Croatian. English transcription works everywhere, but smaller languages are exactly where speech recognition models are weakest, and Central and Southeast Europe is our core market. We felt it where it hurts most, in a live sales demo, when the Serbian transcript came out visibly weaker than the product deserved. When your transcript is wrong, everything downstream is wrong, the summary, the tasks, the credibility. We handled it on three levels. First, we self-host Whisper, so audio never leaves our servers and we control the model instead of renting someone else's black box. Second, we benchmark word error rate and character error rate separately for Serbian and Croatian because aggregate multilingual numbers hide exactly the languages our customers speak. Third, we exposed a Speed versus Quality slider, so the user chooses the tradeoff for each meeting instead of us guessing for them. The complaint made the product better, which is what complaints are for.

Most people think transcription is the product. It is not; it is a commodity, and it gets cheaper every quarter. The product is accountability. Who decided what, who owns which task, by when, with named responsibility. A perfect transcript of a meeting where nobody owns anything is a perfect record of wasted time. The second thing people get wrong is the region itself. The big US tools treat Central and Southeast Europe as a rounding error. Languages like Albanian, Macedonian, Slovenian, and Serbian are afterthoughts in their roadmaps, and EU data residency is a compliance checkbox they hope nobody asks about. For a Hungarian or Serbian SME, those two things are the entire buying decision. Their meetings happen in their language, and their lawyers ask where the audio physically sits. So we generate output in 13 regional languages and keep everything inside the EEA. What looks like a niche from San Francisco looks like an underserved home market from Belgrade.

Scaling the product down, not up. The goal is that everything Minuteory does gets delivered from one button on your screen. Record, and the summary, the decisions, and the named tasks land where your team already works, pushed straight into Microsoft Planner and Teams instead of dying in a document nobody opens. On top of that, we are adding goals and phases, so meetings stop being isolated events and start rolling up into something measurable, tasks connect to phases, phases to goals, and you can track OKRs and KPIs and keep project delivery aligned without a separate tool. A PARA Notes feature that converts loose observations into accountable tasks, and a second extraction pass to raise task recall. But the customer should see none of that. One button, and the meeting is handled and accounted for.

We are happy to handle more than 10,000 tasks for a client. Created, related, and reviewed from meetings in Minuteory.

Not from scratch on operations, completely from scratch on selling my own SaaS. I spent six years as a technical product manager at Samsung Electronics Adria, launching mobile portfolios across the region, then worked as a business analyst at Endava on SaaS implementations for a US communications company. After that, I headed HYCU's Belgrade delivery center, scaling it from 18 to 65 people while building product and service teams and opening support and inside sales functions from zero. Most recently, I was Director of Technology at LoginEKO, where I restructured a 160-person organization twice. Twenty years inside the exact meetings this product now processes. What I had never done before was find customers, price a product, and sell it myself.

We built our website as a React single-page application with translations fetched at runtime from the database. It looked great to humans. To Google and AI assistants, it was invisible because crawlers saw placeholder keys instead of actual content. Every route served the same meta tags and the same canonical URL. The result, barely indexed in search and completely absent when AI assistants recommend tools in our own category, which is an increasingly common way buyers discover software. We were an AI product that AI could not see. The lesson cost us months of organic traffic, SEO is architecture, not content. You decide it on day one, in the framework choice, before writing a single blog post.